image.png

Important Python Libraries¶

In [1]:
##---To remove warnings - make code output readable---#
import warnings
warnings.filterwarnings("ignore")

##---For DataFrame manipulation and Numpy array--#
import pandas as pd
import numpy as np

##---For visualization---#
import matplotlib.pyplot as plt
plt.style.use('seaborn-dark-palette')
import seaborn as sns
from IPython.display import display, HTML

# library for Machine Learning Models
from sklearn.ensemble import RandomForestClassifier, RandomForestRegressor
from sklearn.ensemble import GradientBoostingClassifier, GradientBoostingRegressor

# Library for Training and testing data split
from sklearn.model_selection import train_test_split

# libraries for confusion Matrix, Classification Repports, ROC and AUC, R2, MAE, RMSE, MSE
from sklearn.metrics import confusion_matrix, classification_report, roc_curve, auc
from sklearn.metrics import mean_squared_error, mean_absolute_error,r2_score

Load Data¶

In [ ]:
data = pd.read_excel("Book1DATA.xlsx")
In [3]:
print("There are total {} loan Applicants".format(data["ApplicantId"].nunique()))

There are total 75748 loan Applicants

Remove those columns where all values are Missing/ Null/ NaN

In [4]:
columns_to_consider = []
for i in data.columns:
    if data[i].isna().sum() != 75748:
        columns_to_consider.append(i)
        
print("There are {} columns/ potential features".format(data.shape[1]))
print("There are {} columns where all values are Null".format(data.shape[1] - len(columns_to_consider)))

There are 171 columns/ potential features
There are 26 columns where all values are Null
In [141]:
sub_dataset = data[columns_to_consider]

Data Pre-processing¶

  • Convert categories such as Final Expired 0/1/2 to Final Expired
In [142]:
sub_dataset["LoanStatus"] = sub_dataset["LoanStatus"].replace("Final - Expired 0", "Final - Expired")
sub_dataset["LoanStatus"] = sub_dataset["LoanStatus"].replace("Final - Expired 1", "Final - Expired")
sub_dataset["LoanStatus"] = sub_dataset["LoanStatus"].replace("Final - Expired 2", "Final - Expired")
sub_dataset["LoanStatus"] = sub_dataset["LoanStatus"].replace("Final - Expired 3", "Final - Expired")
sub_dataset["LoanStatus"] = sub_dataset["LoanStatus"].replace("Final - Expired 4", "Final - Expired")

sub_dataset["LoanStatus"] = sub_dataset["LoanStatus"].replace("Final - Canceled 0", "Final - Canceled")
sub_dataset["LoanStatus"] = sub_dataset["LoanStatus"].replace("Final - Canceled 1", "Final - Canceled")
sub_dataset["LoanStatus"] = sub_dataset["LoanStatus"].replace("Final - Canceled 2", "Final - Canceled")
sub_dataset["LoanStatus"] = sub_dataset["LoanStatus"].replace("Final - Canceled 3", "Final - Canceled")
sub_dataset["LoanStatus"] = sub_dataset["LoanStatus"].replace("Final - Canceled 4", "Final - Canceled")

sub_dataset["AUSStatus"].value_counts(normalize=True).mul(100).round(2)
Out[142]:
Open                      61.83
Denied                    19.23
Approved                  11.87
In Review                  6.49
Conditionally Approved     0.59
Name: AUSStatus, dtype: float64
In [143]:
sub_dataset["Mailing State"] = sub_dataset["Mailing State"].fillna(sub_dataset["Mailing State"].mode())
sub_dataset["EmploymentType"] = sub_dataset["EmploymentType"].fillna(sub_dataset["EmploymentType"].mode())
sub_dataset["ProjectType"] = sub_dataset["ProjectType"].fillna("DidNotMentioned")

Potential Features¶

In [144]:
features = ["ApplicantId","Mailing State", "EmploymentType", "AgeAtApplication", "StatedMonthlyIncome", "FrozenBureaus", "XPNFICO",\
           "TUFICO", "XPNBK", "FinancialProduct", "ApplicationDateDim", "ProjectType", "Property State", "RequestedAmount",\
           "DecisionFICOScore", "DecisionBKScore", "AUSStatus", "AutoDecision", "ApplicationDate","AutoDecisionDate", \
            "SoftDecision", "HardDecision", "MaxOfferAmount", "LoanId","LoanStatus", "LoanStage", "LoanAmount", \
            "LoanTermInMonths", "MonthlyPayment", "Program",\
            "APR","InterestRate", 'FundedDateDim', 'NetFundedAmount', "FinancingPlan", "TotalFinanceCharge", \
            "GrossDealerFee", "GrossDealerFeeRate", "DiscountDealerFee", "DiscountDealerFeeRate","NetDealerFee",\
            "NetDealerFeeRate", "UWMonthlyDebtPayments", "UWPreLoanDTI", "UWPostLoanDTI",\
            "ContractorOrganizationId", "ContractorId", "Contrator User First Name", "Contrator User Last Name",\
            "Contrator User Email", "Contrator User Phone", "ContractorKey", "ContractorJobNumber", "ContractorNumber",\
            "HEROAccountManagerId", "ContractorOrganizationName", "ContractorCity", "LocationNickname", "Branch",\
            "ContractorTypeName", "BoostList", "Contractor Applied Date", "Contractor Registration Date", \
            "AccountStatus", "AccountStatusName", "Contractor Effective Date", \
            "BenjiDealerFeeDiscountRate", "HERODealerFeeDiscountRate", "UserId", "UserName"]

sub_dataset = sub_dataset[features]

Missing Values Analysis¶

In [145]:
sub_dataset.isna().sum()
Out[145]:
ApplicantId                         0
Mailing State                       4
EmploymentType                      8
AgeAtApplication                    0
StatedMonthlyIncome              2675
FrozenBureaus                   42668
XPNFICO                         25851
TUFICO                          70994
XPNBK                           25266
FinancialProduct                    0
ApplicationDateDim                  0
ProjectType                         0
Property State                      0
RequestedAmount                 69734
DecisionFICOScore               20045
DecisionBKScore                 38503
AUSStatus                           0
AutoDecision                       53
ApplicationDate                     0
AutoDecisionDate                   42
SoftDecision                     3408
HardDecision                    53614
MaxOfferAmount                  39002
LoanId                              0
LoanStatus                          0
LoanStage                           0
LoanAmount                      51559
LoanTermInMonths                51559
MonthlyPayment                  51559
Program                         23361
                                ...  
NetDealerFee                    54669
NetDealerFeeRate                54669
UWMonthlyDebtPayments           24776
UWPreLoanDTI                    24787
UWPostLoanDTI                   43547
ContractorOrganizationId            0
ContractorId                        0
Contrator User First Name           0
Contrator User Last Name            0
Contrator User Email                0
Contrator User Phone                1
ContractorKey                       0
ContractorJobNumber             71095
ContractorNumber                    0
HEROAccountManagerId                0
ContractorOrganizationName          0
ContractorCity                      0
LocationNickname                    0
Branch                          25202
ContractorTypeName              18252
BoostList                           0
Contractor Applied Date             0
Contractor Registration Date        0
AccountStatus                       0
AccountStatusName                   0
Contractor Effective Date           0
BenjiDealerFeeDiscountRate       2030
HERODealerFeeDiscountRate           0
UserId                              0
UserName                         1095
Length: 70, dtype: int64
In [146]:
sub_dataset["FirstLastNameAddress"] = data["FirstName"].astype(str)+" "+data["LastName"].astype(str)+" "+data["Mailing Address1"].astype(str)

temp = data[["ApplicantId"]]
temp["ApplicantName"] = data["FirstName"].astype(str)+" "+data["LastName"].astype(str)+" "+data["Mailing Address1"].astype(str)
temp = temp.groupby("ApplicantName").ApplicantId.nunique().reset_index()
temp.columns = ["FirstLastNameAddress", "WhetherApplicantAvailLoanPreviously"]

sub_dataset = sub_dataset.merge(temp, on="FirstLastNameAddress", how="left")
In [147]:
sub_dataset.to_csv("Processed_Data.csv", index=False)
In [148]:
data['days_to_go'] = (data.DocsSentDate - data.ApplicationDate).dt.days
data[(~data['days_to_go'].isna()) & (data['days_to_go'] !=0 )][['days_to_go']].mean()
Out[148]:
days_to_go    14.199177
dtype: float64

Exploratory Data Analysis¶

In [90]:
ignore = ["ApplicantId","ApplicationDateDim","ApplicationDate","AutoDecisionDate", \
            "LoanId",'FundedDateDim', "ContractorOrganizationId", "Contrator User First Name",\
             "Contrator User Last Name","Contrator User Email", "Contrator User Phone", "ContractorKey", "ContractorJobNumber",\
            "ContractorNumber", "HEROAccountManagerId", "ContractorOrganizationName","ContractorTypeName",\
            "Contractor Applied Date", "Contractor Registration Date", "AccountStatusName", \
            "Contractor Effective Date", "UserId", "UserName"]
for i in sub_dataset.columns:
    
    if i in ignore:
        continue
        
    plt.figure(figsize=(20, 6))
    sub_dataset[i].value_counts().head(20).plot.bar()
    plt.title(i)
    plt.show()
    
    temp = sub_dataset[i].value_counts(normalize=True).mul(100).head(20).reset_index()
    temp[i] = round(temp[i], 2)
    temp[i] = temp[i].astype(str)+"%"
    display(temp)
index Mailing State
0 CA 19.92%
1 FL 12.5%
2 TX 12.08%
3 AZ 4.26%
4 MO 3.98%
5 IL 3.54%
6 GA 3.19%
7 NC 3.0%
8 OH 2.77%
9 KS 2.39%
10 TN 2.18%
11 NY 2.03%
12 SC 1.95%
13 MI 1.91%
14 CO 1.87%
15 MD 1.85%
16 IA 1.84%
17 IN 1.7%
18 AL 1.57%
19 VA 1.52%
index EmploymentType
0 Employed 70.94%
1 Retired 21.78%
2 SelfEmployed 5.52%
3 NotEmployed 1.75%
4 0 0.01%
index AgeAtApplication
0 58 2.39%
1 59 2.37%
2 55 2.33%
3 57 2.32%
4 60 2.3%
5 62 2.29%
6 63 2.29%
7 50 2.28%
8 56 2.28%
9 61 2.23%
10 51 2.2%
11 54 2.17%
12 52 2.12%
13 64 2.11%
14 49 2.11%
15 53 2.01%
16 66 2.0%
17 48 1.98%
18 47 1.97%
19 65 1.96%
index StatedMonthlyIncome
0 0.0 9.18%
1 5000.0 4.19%
2 10000.0 3.53%
3 6000.0 3.4%
4 4000.0 2.92%
5 8000.0 2.8%
6 7000.0 2.31%
7 7500.0 2.12%
8 3000.0 1.94%
9 4500.0 1.72%
10 9000.0 1.59%
11 3500.0 1.57%
12 6500.0 1.53%
13 12000.0 1.46%
14 5500.0 1.36%
15 2500.0 1.27%
16 15000.0 1.16%
17 8500.0 1.03%
18 2000.0 0.92%
19 20000.0 0.86%
index FinancialProduct
0 Benji 68.13%
1 LendingPoint 31.87%
index ProjectType
0 DidNotMentioned 67.13%
1 HVAC 12.28%
2 Multi-Product 6.36%
3 Roofing 3.36%
4 Windows 2.43%
5 General Plumbing 1.32%
6 Insulation 1.03%
7 Landscaping 0.72%
8 Cabinets/Shelving 0.67%
9 Interior Flooring 0.58%
10 Bathroom Remodeling 0.54%
11 Siding 0.52%
12 Kitchen Remodeling 0.51%
13 Doors 0.27%
14 Foundation 0.25%
15 Patio Covers 0.23%
16 Paint - Exterior 0.22%
17 Solar 0.19%
18 Water Proofing 0.11%
19 Carports 0.11%
index Property State
0 CA 19.92%
1 FL 12.55%
2 TX 12.07%
3 AZ 4.29%
4 MO 4.01%
5 IL 3.54%
6 GA 3.19%
7 NC 2.99%
8 OH 2.76%
9 KS 2.37%
10 TN 2.18%
11 NY 2.0%
12 SC 1.96%
13 MI 1.92%
14 CO 1.87%
15 MD 1.85%
16 IA 1.84%
17 IN 1.7%
18 AL 1.57%
19 VA 1.52%
index RequestedAmount
0 0.0 92.06%
1 25000.0 0.31%
2 5000.0 0.13%
3 10000.0 0.11%
4 20000.0 0.09%
5 4000.0 0.06%
6 17500.0 0.06%
7 7000.0 0.05%
8 12650.0 0.05%
9 9000.0 0.05%
10 15000.0 0.05%
11 8500.0 0.04%
12 6500.0 0.04%
13 11000.0 0.04%
14 8000.0 0.04%
15 7500.0 0.04%
16 2000.0 0.03%
17 6000.0 0.03%
18 13000.0 0.03%
19 3000.0 0.03%
index LoanTermInMonths
0 0.0 68.07%
1 60.0 14.88%
2 120.0 6.71%
3 84.0 4.71%
4 144.0 3.15%
5 72.0 1.01%
6 48.0 0.63%
7 36.0 0.53%
8 24.0 0.22%
9 180.0 0.09%
index WhetherApplicantAvailLoanPreviously
0 2 53.44%
1 1 37.22%
2 4 4.93%
3 3 3.15%
4 6 0.58%
5 5 0.51%
6 7 0.06%
7 8 0.05%
8 10 0.03%
9 12 0.02%
10 11 0.01%
index MonthlyPayment
0 0.00 68.07%
1 167.10 0.12%
2 851.59 0.12%
3 264.19 0.11%
4 238.90 0.11%
5 330.24 0.11%
6 396.29 0.11%
7 359.09 0.09%
8 358.35 0.08%
9 298.63 0.08%
10 272.26 0.06%
11 896.56 0.06%
12 198.80 0.05%
13 53.78 0.05%
14 56.78 0.05%
15 567.73 0.05%
16 806.62 0.05%
17 404.04 0.05%
18 587.55 0.04%
19 283.86 0.04%
index InterestRate
0 0.0000 68.07%
1 0.0999 10.51%
2 0.1799 9.58%
3 0.0699 2.51%
4 0.2399 2.31%
5 0.0899 1.12%
6 0.1099 0.91%
7 0.1431 0.6%
8 0.1199 0.52%
9 0.1299 0.38%
10 0.1631 0.38%
11 0.1331 0.37%
12 0.1399 0.36%
13 0.1599 0.26%
14 0.1531 0.24%
15 0.1418 0.24%
16 0.1318 0.23%
17 0.1618 0.18%
18 0.1731 0.17%
19 0.2030 0.12%
index AUSStatus
0 Open 61.83%
1 Denied 19.23%
2 Approved 11.87%
3 In Review 6.49%
4 Conditionally Approved 0.59%
index LoanStatus
0 Final - Denied 52.02%
1 Approved 25.9%
2 Final - Expired 11.49%
3 Final - Canceled 8.66%
4 Preapproved 1.65%
5 In Review 0.22%
6 Conditionally Approved 0.04%
7 Submitted 0.01%
8 Terminated 0.0%
index AutoDecision
0 Denied 33.99%
1 Preapproved 29.3%
2 InReview 27.97%
3 Approved 8.19%
4 ConditionallyApproved 0.48%
5 0 0.07%
index SoftDecision
0 Denied 51.91%
1 Preapproved 31.7%
2 Approved 11.88%
3 0 4.5%
4 3 0.01%
index HardDecision
0 0 70.78%
1 Approved 29.0%
2 Denied 0.22%
index XPNFICO
0 0.0 34.13%
1 850.0 0.78%
2 685.0 0.32%
3 694.0 0.31%
4 697.0 0.3%
5 686.0 0.3%
6 774.0 0.3%
7 696.0 0.29%
8 736.0 0.29%
9 826.0 0.28%
10 723.0 0.28%
11 669.0 0.28%
12 710.0 0.28%
13 704.0 0.27%
14 652.0 0.27%
15 813.0 0.27%
16 718.0 0.27%
17 674.0 0.27%
18 787.0 0.27%
19 700.0 0.27%
index TUFICO
0 0.0 93.72%
1 850.0 0.1%
2 738.0 0.05%
3 829.0 0.04%
4 687.0 0.04%
5 715.0 0.04%
6 670.0 0.04%
7 671.0 0.04%
8 726.0 0.04%
9 741.0 0.04%
10 685.0 0.04%
11 677.0 0.04%
12 694.0 0.04%
13 658.0 0.04%
14 662.0 0.04%
15 678.0 0.03%
16 720.0 0.03%
17 676.0 0.03%
18 704.0 0.03%
19 686.0 0.03%
index XPNBK
0 0.0 33.36%
1 529.0 0.27%
2 628.0 0.26%
3 491.0 0.25%
4 507.0 0.25%
5 595.0 0.25%
6 601.0 0.24%
7 518.0 0.24%
8 647.0 0.24%
9 543.0 0.24%
10 610.0 0.24%
11 438.0 0.24%
12 515.0 0.24%
13 557.0 0.23%
14 462.0 0.23%
15 537.0 0.23%
16 657.0 0.23%
17 499.0 0.23%
18 540.0 0.23%
19 480.0 0.23%
index DecisionFICOScore
0 0.0 27.86%
1 850.0 0.92%
2 694.0 0.35%
3 618.0 0.34%
4 686.0 0.34%
5 685.0 0.33%
6 657.0 0.32%
7 696.0 0.32%
8 697.0 0.31%
9 736.0 0.31%
10 669.0 0.31%
11 704.0 0.31%
12 674.0 0.31%
13 700.0 0.31%
14 718.0 0.3%
15 676.0 0.3%
16 826.0 0.3%
17 646.0 0.3%
18 658.0 0.3%
19 622.0 0.3%
index DecisionBKScore
0 0.0 51.87%
1 520.0 0.16%
2 103.0 0.16%
3 438.0 0.16%
4 647.0 0.15%
5 518.0 0.15%
6 557.0 0.15%
7 480.0 0.15%
8 407.0 0.15%
9 543.0 0.15%
10 537.0 0.15%
11 507.0 0.15%
12 450.0 0.15%
13 601.0 0.15%
14 462.0 0.15%
15 499.0 0.14%
16 529.0 0.14%
17 580.0 0.14%
18 447.0 0.14%
19 452.0 0.14%
index Program
0 0 30.84%
1 RICCA 13.34%
2 PROMFL 8.6%
3 RICTX 6.45%
4 RICMO 2.52%
5 RICAZGBRAZIL 1.77%
6 RICKS 1.69%
7 RICIL 1.48%
8 RICNC 1.44%
9 PROMGA 1.44%
10 RICAZ 1.36%
11 RICIA 1.28%
12 PROMOH 1.27%
13 RICCO 1.22%
14 RICNE 1.21%
15 PROMNYSEARS 1.16%
16 RICSC 1.11%
17 RICTXSEARS 1.07%
18 RICIN 0.99%
19 PROMTN 0.99%
index APR
0 0.0000 68.07%
1 0.0999 10.43%
2 0.1799 9.58%
3 0.0699 2.51%
4 0.0899 1.12%
5 0.1431 0.92%
6 0.1099 0.91%
7 0.2547 0.62%
8 0.1533 0.57%
9 0.1735 0.53%
10 0.1199 0.52%
11 0.2575 0.49%
12 0.1532 0.44%
13 0.2621 0.43%
14 0.2515 0.3%
15 0.1610 0.21%
16 0.2493 0.19%
17 0.2140 0.18%
18 0.1736 0.18%
19 0.1813 0.17%
index BoostList
0 No 90.38%
1 Yes 9.62%
index BenjiDealerFeeDiscountRate
0 2.0 42.81%
1 0.0 40.12%
2 1.0 7.7%
3 2.5 4.26%
4 1.5 3.75%
5 0.5 1.36%
index HERODealerFeeDiscountRate
0 0 71.71%
1 5 20.2%
2 4 4.72%
3 2 1.87%
4 1 0.76%
5 3 0.51%
6 10 0.1%
7 6 0.08%
8 7 0.05%
In [12]:
other_features = ["XPNFICO", "TUFICO", "XPNBK",\
                 "DecisionFICOScore", "DecisionBKScore", "Program", "APR", "BoostList",\
                 "BenjiDealerFeeDiscountRate", "HERODealerFeeDiscountRate"]

Make Data¶

In [78]:
columns = ["Mailing State", "EmploymentType", "AgeAtApplication", "StatedMonthlyIncome", "FinancialProduct",\
            "ApplicationDateDim", "ProjectType", "Property State", "RequestedAmount", "LoanTermInMonths", \
           "WhetherApplicantAvailLoanPreviously",
            "MonthlyPayment", "InterestRate", "AUSStatus", "LoanStatus", "AutoDecision", "SoftDecision", "HardDecision"]

columns.extend(other_features)

sub_dataset = sub_dataset.fillna(0.0)

sub_dataset = sub_dataset[columns]

Label Encoding¶

In [14]:
for i in ["Mailing State", "EmploymentType", "FinancialProduct", "ApplicationDateDim", "ProjectType", "Property State", \
         "Program", "BoostList"]:
    LAYER = pd.get_dummies(sub_dataset[i], prefix=i+"_")
    sub_dataset[LAYER.columns] = LAYER
    sub_dataset = sub_dataset.drop(columns=[i])
sub_dataset.head(5)
Out[14]:
AgeAtApplication StatedMonthlyIncome RequestedAmount LoanTermInMonths WhetherApplicantAvailLoanPreviously MonthlyPayment InterestRate AUSStatus LoanStatus AutoDecision ... Program__RICWA Program__RICWASEARS Program__RICWI Program__RICWISEARS Program__RICWV Program__RICWVSEARS Program__RICWY Program__RICWaSEARS BoostList__No BoostList__Yes
0 70 14560.0 0.0 60.0 1 127.22 0.1799 Approved Approved Approved ... 0 0 0 0 0 0 0 0 1 0
1 72 7600.0 0.0 0.0 2 0.00 0.0000 Denied Final - Denied Denied ... 0 0 0 0 0 0 0 0 1 0
2 56 12000.0 0.0 60.0 1 631.77 0.1799 Approved Approved Approved ... 0 0 0 0 0 0 0 0 1 0
3 32 4400.0 0.0 0.0 2 0.00 0.0000 Open Final - Denied InReview ... 0 0 0 0 0 0 0 0 1 0
4 35 7100.0 0.0 60.0 1 268.87 0.1799 Open Approved Preapproved ... 0 0 0 0 0 0 0 0 1 0

5 rows × 354 columns

In [15]:
dataset_ausstaus = sub_dataset[sub_dataset["AUSStatus"].isin(["Denied", "Approved"])][sub_dataset.columns]
Y_ausstatus = sub_dataset[sub_dataset["AUSStatus"].isin(["Denied", "Approved"])][["AUSStatus"]]
dataset_ausstaus = dataset_ausstaus.drop(columns=["AUSStatus", "LoanStatus", "AutoDecision", "SoftDecision", "HardDecision"])

dataset_loan = sub_dataset[sub_dataset["LoanStatus"].isin(["Final - Denied","Approved", "Final - Expired", "Final - Canceled"])][sub_dataset.columns]
Y_loan_status = sub_dataset[sub_dataset["LoanStatus"].isin(["Final - Denied","Approved", "Final - Expired", "Final - Canceled"])][["LoanStatus"]]
dataset_loan = dataset_loan.drop(columns=["AUSStatus", "LoanStatus", "AutoDecision", "SoftDecision", "HardDecision"])

dataset_auto_decision = sub_dataset[sub_dataset["AutoDecision"].isin(["Denied", "Approved"])][sub_dataset.columns]
Y_auto_decision = sub_dataset[sub_dataset["AutoDecision"].isin(["Denied", "Approved"])][["AutoDecision"]]
dataset_auto_decision = dataset_auto_decision.drop(columns=["AUSStatus", "LoanStatus", "AutoDecision", "SoftDecision", "HardDecision"])

dataset_soft_decision = sub_dataset[sub_dataset["SoftDecision"].isin(["Denied", "Approved"])][sub_dataset.columns]
Y_soft_decision = sub_dataset[sub_dataset["SoftDecision"].isin(["Denied", "Approved"])][["SoftDecision"]]
dataset_soft_decision = dataset_soft_decision.drop(columns=["AUSStatus", "LoanStatus", "AutoDecision", "SoftDecision", "HardDecision"])

dataset_hard_decision = sub_dataset[sub_dataset["HardDecision"].isin(["Denied", "Approved"])][sub_dataset.columns]
Y_hard_decision = sub_dataset[sub_dataset["HardDecision"].isin(["Denied", "Approved"])][["HardDecision"]]
dataset_hard_decision = dataset_hard_decision.drop(columns=["AUSStatus", "LoanStatus", "AutoDecision", "SoftDecision", "HardDecision"])

Machine Learning Technique¶

In [151]:
def trainMLAlgorithm(X, Y):    
    X_train, X_test, y_train, y_test = train_test_split(X, Y, test_size=0.30, stratify=Y)
    
    # make Random Forest Classifier object
    forest = RandomForestClassifier(n_estimators=500,random_state=1)
    
    # Train Algo on training data
    forest.fit(X_train, y_train)
    
    # Predict target for testing data
    y_pred = forest.predict(X_test)
    
    print("Classification Report")
    confusion_report = classification_report(y_test, y_pred)
    print(confusion_report)
    
    print("\nConfusion Matrix")
    confusion_attack = confusion_matrix(y_test, y_pred)
    print(confusion_attack)
    
    plt.figure(figsize=(18,8))
    feat_impts = [] 
    # calculating features
    feat_impts.append(forest.feature_importances_)
    # ploting features using
    (pd.Series(np.mean(feat_impts, axis=0), index=X.columns)
       .nlargest(49)
       .plot(kind='bar'))
    
    plt.show()

Loan Approval Detection

image.png

AUSStatus Detection using Machine Learning Based on Features¶

In [17]:
trainMLAlgorithm(dataset_ausstaus, Y_ausstatus)

Classification Report
              precision    recall  f1-score   support

    Approved       0.99      0.98      0.98      2697
      Denied       0.98      0.99      0.99      4369

    accuracy                           0.99      7066
   macro avg       0.99      0.99      0.99      7066
weighted avg       0.99      0.99      0.99      7066


Confusion Matrix
[[2630   67]
 [  22 4347]]

LoanStatus Detection using Machine Learning Based on Features¶

In [18]:
trainMLAlgorithm(dataset_loan, Y_loan_status)

Classification Report
                  precision    recall  f1-score   support

        Approved       0.92      0.99      0.95      5886
Final - Canceled       0.58      0.41      0.48      1969
  Final - Denied       0.90      0.95      0.93     11821
 Final - Expired       0.63      0.48      0.54      2612

        accuracy                           0.86     22288
       macro avg       0.76      0.71      0.73     22288
    weighted avg       0.84      0.86      0.85     22288


Confusion Matrix
[[ 5844    12     0    30]
 [  178   816   574   401]
 [   24   208 11287   302]
 [  319   364   683  1246]]

AutoDecision Detection using Machine Learning Based on Features¶

In [19]:
trainMLAlgorithm(dataset_auto_decision, Y_auto_decision)

Classification Report
              precision    recall  f1-score   support

    Approved       0.99      0.98      0.98      1861
      Denied       0.99      1.00      1.00      7724

    accuracy                           0.99      9585
   macro avg       0.99      0.99      0.99      9585
weighted avg       0.99      0.99      0.99      9585


Confusion Matrix
[[1821   40]
 [  20 7704]]

SoftDecision Detection using Machine Learning Based on Features¶

In [20]:
trainMLAlgorithm(dataset_soft_decision, Y_soft_decision)

Classification Report
              precision    recall  f1-score   support

    Approved       0.99      0.97      0.98      2700
      Denied       0.99      1.00      1.00     11796

    accuracy                           0.99     14496
   macro avg       0.99      0.98      0.99     14496
weighted avg       0.99      0.99      0.99     14496


Confusion Matrix
[[ 2619    81]
 [   35 11761]]

HardDecision Detection using Machine Learning Based on Features¶

In [21]:
trainMLAlgorithm(dataset_hard_decision, Y_hard_decision)

Classification Report
              precision    recall  f1-score   support

    Approved       0.99      1.00      1.00      6591
      Denied       0.94      0.30      0.45        50

    accuracy                           0.99      6641
   macro avg       0.97      0.65      0.73      6641
weighted avg       0.99      0.99      0.99      6641


Confusion Matrix
[[6590    1]
 [  35   15]]

image.png

Loan Approval Duration Prediction¶

  • In how many days after application submition, loan can be approved or rejected
In [153]:
duration = data[data["LoanStatus"] == "Approved"][["ApplicantId","ApplicationDate", "HardDecisionDate"]]
duration = duration.dropna()
duration["DaysToApproveLoan"] = (pd.to_datetime(duration["HardDecisionDate"]) - pd.to_datetime(duration["ApplicationDate"])).dt.days
duration.loc[duration["DaysToApproveLoan"] == 0, "DaysToApproveLoan"] = 1
In [154]:
dataset_loan_reg = dataset_loan.copy()
dataset_loan_reg["ApplicantId"] = data["ApplicantId"]
dataset_loan_reg = dataset_loan_reg.merge(duration[["ApplicantId", "DaysToApproveLoan"]], on="ApplicantId", how="left")
dataset_loan_reg["DaysToApproveLoan"] = dataset_loan_reg["DaysToApproveLoan"].fillna(0)
In [155]:
Y_time = dataset_loan_reg[["DaysToApproveLoan"]]
X = dataset_loan_reg.drop(columns=["ApplicantId", "DaysToApproveLoan"])
In [157]:
X_train, X_test, y_train, y_test = train_test_split(X, Y_time, test_size=0.30)

# make Random Forest Classifier object
model = RandomForestRegressor(n_estimators=500, random_state=1)

# Train Algo on training data
model.fit(X_train, y_train)

# Predict target for testing data
y_pred = model.predict(X_test)

import math
mse = mean_squared_error(y_test, y_pred)
print('MSE: '+str(mse))
mae = mean_absolute_error(y_test, y_pred)
print('MAE: '+str(mae))
rmse = math.sqrt(mean_squared_error(y_test, y_pred))
print('RMSE: '+str(rmse))
rsqrt = r2_score(y_test, y_pred)
print('R-Squared: '+str(rsqrt))

MSE: 82.78938213020636
MAE: 2.9586823049500435
RMSE: 9.098867079488873
R-Squared: -0.2564216174639844

---------------------------------------------------------------------------
NameError                                 Traceback (most recent call last)
<ipython-input-157-4cce3dd36790> in <module>
     20 print('R-Squared: '+str(rsqrt))
     21 
---> 22 (pd.Series(np.mean(feat_impts, axis=0), index=X.columns)
     23        .nlargest(49)
     24        .plot(kind='bar'))

NameError: name 'feat_impts' is not defined
In [158]:
plt.figure(figsize=(18,8))
feat_impts = [] 

# calculating features
feat_impts.append(model.feature_importances_)
(pd.Series(np.mean(feat_impts, axis=0), index=X.columns)
       .nlargest(49)
       .plot(kind='bar'))  
plt.show()
    
    
fig = plt.figure(figsize=(16,8))
a = [i for i in range(len(y_test))]
L = 0
H = 100
plt.plot(a[L:H], y_test[L:H], label='Actual Days to Approve Loan')
plt.plot(a[L:H], y_pred[L:H], label='Predicted Days to Approve Loan')
plt.legend()
plt.ylabel("Days to Approve Loan")
plt.show()
In [159]:
fig = plt.figure(figsize=(16,8))
a = [i for i in range(len(y_test))]
L = 100
H = 500
plt.plot(a[L:H], y_test[L:H], label='Actual Days to Approve Loan')
plt.plot(a[L:H], y_pred[L:H], label='Predicted Days to Approve Loan')
plt.legend()
plt.ylabel("Days to Approve Loan")
plt.show()
In [167]:
dataset_loan_reg["DaysToApproveLoan"].value_counts()
Out[167]:
0.0      55073
1.0      12119
2.0       1143
3.0        859
4.0        586
5.0        501
6.0        461
7.0        340
8.0        245
9.0        196
13.0       173
10.0       173
14.0       171
12.0       162
15.0       144
11.0       141
21.0       119
16.0       106
20.0       106
17.0       100
18.0        84
28.0        77
29.0        70
19.0        67
25.0        65
27.0        62
26.0        59
22.0        58
23.0        57
24.0        49
         ...  
155.0        1
121.0        1
128.0        1
157.0        1
159.0        1
170.0        1
156.0        1
164.0        1
202.0        1
178.0        1
192.0        1
150.0        1
188.0        1
196.0        1
237.0        1
216.0        1
208.0        1
205.0        1
149.0        1
145.0        1
225.0        1
132.0        1
116.0        1
165.0        1
123.0        1
101.0        1
112.0        1
213.0        1
190.0        1
142.0        1
Name: DaysToApproveLoan, Length: 171, dtype: int64
In [181]:
# dataset_loan_reg[dataset_loan_reg["FinancialProduct__Benji"] == 1].mean()
# dataset_loan_reg[dataset_loan_reg["FinancialProduct__LendingPoint"] == 1].mean()

image.png

How much loan will be funded?¶

In [187]:
loan = data[["ApplicantId","LoanAmount"]]
dataset_loan_amount = dataset_loan_reg.merge(loan, on="ApplicantId", how="left")
dataset_loan_amount["LoanAmount"] = dataset_loan_amount["LoanAmount"].fillna(0)
In [188]:
Y_amount = dataset_loan_amount[["LoanAmount"]]
X = dataset_loan_amount.drop(columns=["ApplicantId", "LoanAmount", "DaysToApproveLoan"])
In [189]:
X_train, X_test, y_train, y_test = train_test_split(X, Y_amount, test_size=0.30)

# make Random Forest Classifier object
model = RandomForestRegressor(n_estimators=500, random_state=1)

# Train Algo on training data
model.fit(X_train, y_train)

# Predict target for testing data
y_pred = model.predict(X_test)

import math
mse = mean_squared_error(y_test, y_pred)
print('MSE: '+str(mse))
mae = mean_absolute_error(y_test, y_pred)
print('MAE: '+str(mae))
rmse = math.sqrt(mean_squared_error(y_test, y_pred))
print('RMSE: '+str(rmse))
rsqrt = r2_score(y_test, y_pred)
print('R-Squared: '+str(rsqrt))

fig = plt.figure(figsize=(16,8))
a = [i for i in range(len(y_test))]
L = 0
H = 100
plt.plot(a[L:H], y_test[L:H], label='Actual Loan Funded')
plt.plot(a[L:H], y_pred[L:H], label='Predicted Loan Funded')
plt.legend()
plt.ylabel("Days to Approve Loan")
plt.show()

MSE: 55380809.465182945
MAE: 5468.475177012533
RMSE: 7441.828368430903
R-Squared: -0.15764032697325892
In [192]:
fig = plt.figure(figsize=(16,8))
a = [i for i in range(len(y_test))]
L = 100
H = 500
plt.plot(a[L:H], y_test[L:H], label='Actual Loan Funded')
plt.plot(a[L:H], y_pred[L:H], label='Predicted Loan Funded')
plt.legend()
plt.ylabel("Loan Amount")
plt.show()
In [191]:
plt.figure(figsize=(18,8))
feat_impts = [] 

# calculating features
feat_impts.append(model.feature_importances_)
(pd.Series(np.mean(feat_impts, axis=0), index=X.columns)
       .nlargest(49)
       .plot(kind='bar'))  
plt.show()

Build models to provide forecast the total number of applications in a time frame like for next months.¶

In [221]:
def nextMonth(model, X):
    next_month = []
    last_rec = list(X[-1])
    new_rec = None
    for i in range(0, 30):
        if new_rec is None:
            new_rec = last_rec
        else:
            new_rec = last_rec[1:]
            new_rec.append(next_month[i-1])
        res = round(model.predict([new_rec])[0])
        next_month.append(res)
        last_rec = new_rec
    return sum(next_month)
In [222]:
data["MonthYear"] = pd.to_datetime(data["ApplicationDate"]).dt.strftime('%Y-%m')
data = data.sort_values(by="MonthYear")
In [223]:
applicant_per_month = data[["ApplicantId", "ApplicationDate"]].groupby(["ApplicationDate"]).count().reset_index()
In [224]:
import numpy
# convert an array of values into a dataset matrix
def create_dataset(dataset, look_back=30):
	dataX, dataY = [], []
	for i in range(len(dataset)-look_back-1):
		a = dataset[i:(i+look_back), 0]
		dataX.append(a)
		dataY.append(dataset[i + look_back, 0])
	return numpy.array(dataX), numpy.array(dataY)
In [225]:
X, y = create_dataset(applicant_per_month["ApplicantId"].values.reshape(-1, 1))
In [226]:
X_train, X_test = X[0:int(0.7*len(X))], X[int(0.7*len(X)):]
y_train, y_test = y[0:int(0.7*len(y))], y[int(0.7*len(y)):]

y_axis = applicant_per_month["ApplicationDate"].values
y_axis = y_axis[int(0.73*len(y_axis)):]

y_axis.shape, y_test.shape
Out[226]:
((88,), (88,))

Linear Regression¶

In [227]:
from sklearn.linear_model import LinearRegression

model = LinearRegression()
model.fit(X_train, y_train)
y_pred = model.predict(X_test)
score = model.score(X_test, y_test)

print('ROC AUC score: {:.2f}'.format(score))

rmse = math.sqrt(mean_squared_error(y_test, y_pred))
print("RMSE = ", rmse)
print("R2 Score = ", r2_score(y_test, y_pred))
print("MAE = ", mean_absolute_error(y_test, y_pred))

plt.figure(figsize=(20,5))
plt.plot(y_axis[0:50], y_test[0:50], '--gx', markersize = 5 , linewidth = 1, label="Actual - Applications for Loan")
plt.plot(y_axis[0:50], y_pred[0:50], '--rx', markersize = 5 , linewidth = 1, label="Predicted - Applications for Loan")
plt.legend()

ROC AUC score: 0.82
RMSE =  39.063634945237304
R2 Score =  0.8168710152000416
MAE =  30.160469258788382
Out[227]:
<matplotlib.legend.Legend at 0x21be5dd2588>

Random Forest Regressor¶

In [228]:
from sklearn.ensemble import RandomForestRegressor
model = RandomForestRegressor(n_estimators=500)
model.fit(X_train, y_train)
y_pred = model.predict(X_test)
score = model.score(X_test, y_test)

print('ROC AUC score: {:.2f}'.format(score))

rmse = math.sqrt(mean_squared_error(y_test, y_pred))
print("RMSE = ", rmse)
print("R2 Score = ", r2_score(y_test, y_pred))
print("MAE = ", mean_absolute_error(y_test, y_pred))

plt.figure(figsize=(20,5))
plt.plot(y_axis[0:50], y_test[0:50], '--gx', markersize = 5 , linewidth = 1, label="Actual - Applications for Loan")
plt.plot(y_axis[0:50], y_pred[0:50], '--rx', markersize = 5 , linewidth = 1, label="Predicted - Applications for Loan")
plt.legend()

ROC AUC score: 0.79
RMSE =  42.162271807861586
R2 Score =  0.786666140833993
MAE =  32.62327272727273
Out[228]:
<matplotlib.legend.Legend at 0x21be622eef0>
In [229]:
fig = plt.figure(figsize=(16,8))
a = [i for i in range(len(y_test))]
L = 0
H = 500
plt.plot(y_axis[L:H], y_test[L:H], label='Actual Loan Funded')
plt.plot(y_axis[L:H], y_pred[L:H], label='Predicted Loan Funded')
plt.legend()
plt.ylabel("Days to Approve Loan")
plt.show()
In [230]:
print("Applications Next Month")
nextMonth(model, X)

Applications Next Month
Out[230]:
6973.0

Build models to provide forecast the total number of applications that will Approve in a time frame like for next months.¶

In [231]:
applicant_approved = data[["ApplicationDate", "LoanStatus"]].groupby(["ApplicationDate"]).LoanStatus.value_counts().unstack(fill_value=0).reset_index()
applicant_approved = applicant_approved[["ApplicationDate","Approved"]]
applicant_approved.head(5)
Out[231]:
LoanStatus ApplicationDate Approved
0 2021-06-01 90
1 2021-06-02 91
2 2021-06-03 100
3 2021-06-04 89
4 2021-06-05 49

Linear Regression¶

In [232]:
X, y = create_dataset(applicant_approved["Approved"].values.reshape(-1, 1))

X_train, X_test = X[0:int(0.7*len(X))], X[int(0.7*len(X)):]
y_train, y_test = y[0:int(0.7*len(y))], y[int(0.7*len(y)):]

y_axis = applicant_per_month["ApplicationDate"].values
y_axis = y_axis[int(0.73*len(y_axis)):]

from sklearn.linear_model import LinearRegression

X_train, X_test, y_train, y_test = train_test_split(X, y, test_size = 0.3, random_state = 3)
model = LinearRegression()
model.fit(X_train, y_train)
y_pred = model.predict(X_test)
score = model.score(X_test, y_test)

print('ROC AUC score: {:.2f}'.format(score))

rmse = math.sqrt(mean_squared_error(y_test, y_pred))
print("RMSE = ", rmse)
print("R2 Score = ", r2_score(y_test, y_pred))
print("MAE = ", mean_absolute_error(y_test, y_pred))

plt.figure(figsize=(20,5))
plt.plot(y_axis[0:50], y_test[0:50], '--gx', markersize = 5 , linewidth = 1, label="Actual - Approved Applications")
plt.plot(y_axis[0:50], y_pred[0:50], '--rx', markersize = 5 , linewidth = 1, label="Predicted - Approved Applications")
plt.legend()

ROC AUC score: 0.80
RMSE =  13.576126580862423
R2 Score =  0.7989501117222652
MAE =  10.547023567273966
Out[232]:
<matplotlib.legend.Legend at 0x21be1a91cc0>

Random Forest Regressor¶

In [233]:
from sklearn.ensemble import RandomForestRegressor
model = RandomForestRegressor()
model.fit(X_train, y_train)
y_pred = model.predict(X_test)
score = model.score(X_test, y_test)

print('ROC AUC score: {:.2f}'.format(score))

rmse = math.sqrt(mean_squared_error(y_test, y_pred))
print("RMSE = ", rmse)
print("R2 Score = ", r2_score(y_test, y_pred))
print("MAE = ", mean_absolute_error(y_test, y_pred))

plt.figure(figsize=(20,5))
plt.plot(y_axis[0:50], y_test[0:50], '--gx', markersize = 5 , linewidth = 1, label="Actual - Approved Applications")
plt.plot(y_axis[0:50], y_pred[0:50], '--rx', markersize = 5 , linewidth = 1, label="Predicted - Approved Applications")
plt.legend()

ROC AUC score: 0.80
RMSE =  13.386893182919968
R2 Score =  0.8045157938225418
MAE =  9.909545454545453
Out[233]:
<matplotlib.legend.Legend at 0x21be3aa5eb8>
In [234]:
fig = plt.figure(figsize=(16,8))
a = [i for i in range(len(y_test))]
L = 0
H = 500
plt.plot(y_axis[L:H], y_test[L:H], label='Actual  - Approved Applications')
plt.plot(y_axis[L:H], y_pred[L:H], label='Predicted  - Approved Applications')
plt.legend()
plt.ylabel("Days to Approve Loan")
plt.show()
In [235]:
print("Applications Next Month")
nextMonth(model, X)

Applications Next Month
Out[235]:
1694.0

Build models to provide forecast the total Loan Amount in a time frame like for next months.¶

In [236]:
loan_approved = data[["ApplicationDate", "LoanAmount"]].groupby(["ApplicationDate"]).sum().reset_index()
loan_approved.head(5)
Out[236]:
ApplicationDate LoanAmount
0 2021-06-01 1305069.27
1 2021-06-02 1474012.56
2 2021-06-03 1469719.40
3 2021-06-04 1375278.22
4 2021-06-05 855077.73
In [237]:
X, y = create_dataset(loan_approved["LoanAmount"].values.reshape(-1, 1))

X_train, X_test = X[0:int(0.7*len(X))], X[int(0.7*len(X)):]
y_train, y_test = y[0:int(0.7*len(y))], y[int(0.7*len(y)):]

y_axis = applicant_per_month["ApplicationDate"].values
y_axis = y_axis[int(0.73*len(y_axis)):]

model = RandomForestRegressor(n_estimators=500)
model.fit(X_train, y_train)
y_pred = model.predict(X_test)
score = model.score(X_test, y_test)

print('ROC AUC score: {:.2f}'.format(score))

rmse = math.sqrt(mean_squared_error(y_test, y_pred))
print("RMSE = ", rmse)
print("R2 Score = ", r2_score(y_test, y_pred))
print("MAE = ", mean_absolute_error(y_test, y_pred))

plt.figure(figsize=(20,5))
plt.plot(y_axis[0:50], y_test[0:50], '--gx', markersize = 5 , linewidth = 1, label="Actual Loan Amount Funded")
plt.plot(y_axis[0:50], y_pred[0:50], '--rx', markersize = 5 , linewidth = 1, label="Predicted Loan Amount Funded")
plt.legend()

ROC AUC score: 0.70
RMSE =  216899.9026781381
R2 Score =  0.6997753067480497
MAE =  169494.00835977282
Out[237]:
<matplotlib.legend.Legend at 0x21be1e92908>
In [238]:
fig = plt.figure(figsize=(16,8))
a = [i for i in range(len(y_test))]
L = 0
H = 500
plt.plot(y_axis[L:H], y_test[L:H], label='Actual  - Loan Amount Funded')
plt.plot(y_axis[L:H], y_pred[L:H], label='Predicted  - Loan Amount Funded')
plt.legend()
plt.ylabel("Days to Approve Loan")
plt.show()
In [241]:
print("Applications Next Month")
nextMonth(model, X)
data["Month"] = pd.to_datetime(data["ApplicationDate"]).dt.month
data[["Month","LoanAmount"]].groupby(["Month"]).sum()

Applications Next Month
Out[241]:
LoanAmount
Month
1 20765264.99
2 21561284.22
3 29534897.14
4 16487268.83
6 38462967.86
7 37097092.89
8 31455291.61
9 27543917.19
10 23847083.12
11 21935927.98
12 19397894.25

Anamoly/ Fraud Detection - Contractor

image.png

Contractor Fraud¶

In this module, we will identify anomalies in contractors, these anomalies may or may not be fraud, but this makes contractor unique from others.

In [46]:
print(features)

['ApplicantId', 'Mailing State', 'EmploymentType', 'AgeAtApplication', 'StatedMonthlyIncome', 'FrozenBureaus', 'XPNFICO', 'TUFICO', 'XPNBK', 'FinancialProduct', 'ApplicationDateDim', 'ProjectType', 'Property State', 'RequestedAmount', 'DecisionFICOScore', 'DecisionBKScore', 'AUSStatus', 'AutoDecision', 'ApplicationDate', 'AutoDecisionDate', 'SoftDecision', 'HardDecision', 'MaxOfferAmount', 'LoanId', 'LoanStatus', 'LoanStage', 'LoanAmount', 'LoanTermInMonths', 'MonthlyPayment', 'Program', 'APR', 'InterestRate', 'FundedDateDim', 'NetFundedAmount', 'FinancingPlan', 'TotalFinanceCharge', 'GrossDealerFee', 'GrossDealerFeeRate', 'DiscountDealerFee', 'DiscountDealerFeeRate', 'NetDealerFee', 'NetDealerFeeRate', 'UWMonthlyDebtPayments', 'UWPreLoanDTI', 'UWPostLoanDTI', 'ContractorOrganizationId', 'ContractorId', 'Contrator User First Name', 'Contrator User Last Name', 'Contrator User Email', 'Contrator User Phone', 'ContractorKey', 'ContractorJobNumber', 'ContractorNumber', 'HEROAccountManagerId', 'ContractorOrganizationName', 'ContractorCity', 'LocationNickname', 'Branch', 'ContractorTypeName', 'BoostList', 'Contractor Applied Date', 'Contractor Registration Date', 'AccountStatus', 'AccountStatusName', 'Contractor Effective Date', 'BenjiDealerFeeDiscountRate', 'HERODealerFeeDiscountRate', 'UserId', 'UserName']
In [91]:
contracter_data = data[["ContractorId"]].drop_duplicates()
temp = data[["ContractorId", "ApplicantId"]].groupby(["ContractorId"])["ApplicantId"].nunique().reset_index()
contracter_data = contracter_data.merge(temp, on="ContractorId", how="left")
contracter_data = contracter_data.rename(columns={"ApplicantId":"TotalApplicants"})
contracter_data.head(5)
Out[91]:
ContractorId TotalApplicants
0 39390 103
1 35197 229
2 36028 9
3 39366 212
4 39358 186
In [92]:
sub_dataset["ContractorId"] = data["ContractorId"]
ContractorId = sub_dataset[["ContractorId", "LoanStatus"]].groupby(["ContractorId"]).LoanStatus.value_counts().unstack(fill_value=0).reset_index()
ContractorId = ContractorId[["ContractorId", "Approved", "Conditionally Approved","Final - Canceled", "Final - Denied"]]
contracter_data = contracter_data.merge(ContractorId, on="ContractorId", how="left")
contracter_data.head(5)
Out[92]:
ContractorId TotalApplicants Approved Conditionally Approved Final - Canceled Final - Denied
0 39390 103 28 0 9 53
1 35197 229 70 0 18 107
2 36028 9 2 0 1 6
3 39366 212 53 0 16 113
4 39358 186 42 0 16 97
In [93]:
sub_dataset["NetFundedAmount"] = data["NetFundedAmount"]
temp = sub_dataset[["ContractorId", "NetFundedAmount"]].groupby(["ContractorId"]).mean().reset_index()
temp.columns = ["ContractorId", "NetFundedAmount_Avg"]
contracter_data = contracter_data.merge(temp, on="ContractorId", how="left")
In [94]:
temp = sub_dataset[sub_dataset["LoanStatus"] == "Approved"][["ContractorId", "NetFundedAmount"]].groupby(["ContractorId"]).min().reset_index()
temp.columns = ["ContractorId", "NetFundedAmount_Minimum"]
temp["NetFundedAmount_Minimum"] = temp["NetFundedAmount_Minimum"].fillna(0.0)
contracter_data = contracter_data.merge(temp, on="ContractorId", how="left")
In [95]:
temp = sub_dataset[["ContractorId", "NetFundedAmount"]].groupby(["ContractorId"]).max().reset_index()
temp.columns = ["ContractorId", "NetFundedAmount_Maximum"]
contracter_data = contracter_data.merge(temp, on="ContractorId", how="left")
In [96]:
temp = sub_dataset[["ContractorId", "NetFundedAmount"]].groupby(["ContractorId"]).sum().reset_index()
temp.columns = ["ContractorId", "NetFundedAmount_Total"]
contracter_data = contracter_data.merge(temp, on="ContractorId", how="left")
In [97]:
sub_dataset["ApplicationDate"] = data["ApplicationDate"]
sub_dataset["ApplicantId"] = data["ApplicantId"]
temp = sub_dataset[["ContractorId", "ApplicationDate", "ApplicantId"]].groupby(["ContractorId", "ApplicationDate"]).count().reset_index()
temp = temp[["ContractorId", "ApplicantId"]].groupby("ContractorId").mean().reset_index()
temp.columns = ["ContractorId", "Average_Applications_per_Day"]
contracter_data = contracter_data.merge(temp, on="ContractorId", how="left")
In [98]:
# Import date class from datetime module
from datetime import date

# Returns the current local date
today = date.today()

sub_dataset["Contractor Registration Date"] = data["Contractor Registration Date"]
temp = sub_dataset[["ContractorId","Contractor Registration Date"]]
temp = temp.drop_duplicates()
temp["TodaysDate"] = today
temp["BusinessAge"] = (pd.to_datetime(temp["TodaysDate"]) - pd.to_datetime(temp["Contractor Registration Date"])).dt.days/30
temp["BusinessAge"] = temp["BusinessAge"].astype(int)
contracter_data = contracter_data.merge(temp[["ContractorId","BusinessAge"]], on="ContractorId", how="left")
In [99]:
contracter_data["Applications_to_Approve_Ratio"] = contracter_data["TotalApplicants"] / contracter_data["Approved"]
contracter_data["Applications_to_Denied_Ratio"] = contracter_data["TotalApplicants"] / contracter_data["Final - Denied"]

contracter_data["Applications_to_Approve_Ratio"] = contracter_data["Applications_to_Approve_Ratio"].replace(math.inf, 0)
contracter_data["Applications_to_Denied_Ratio"] = contracter_data["Applications_to_Denied_Ratio"].replace(math.inf, 0)

contracter_data["Applications_to_Denied_Ratio"] = contracter_data["Applications_to_Denied_Ratio"]
contracter_data["Applications_to_Approve_Ratio"] = contracter_data["Applications_to_Approve_Ratio"]
In [100]:
contracter_data["Over_all_approved_ration"] = contracter_data["Approved"] / contracter_data["BusinessAge"]
contracter_data["Over_all_approved_ration"] = contracter_data["Over_all_approved_ration"].replace(math.inf, 0)
In [101]:
sub_dataset["Mailing State"] = data["Mailing State"]

temp = sub_dataset[["ContractorId", "Mailing State"]].groupby(["ContractorId"])["Mailing State"].nunique().reset_index()
temp.columns = ["ContractorId", "DealsInMoreThanOneState"]
temp.loc[temp["DealsInMoreThanOneState"] == 1, "DealsInMoreThanOneState"] = "No"
temp.loc[~temp["DealsInMoreThanOneState"].isin([1,"No"]), "DealsInMoreThanOneState"] = "Yes"
temp["DealsInMoreThanOneState"] = temp["DealsInMoreThanOneState"].replace("No", 0)
temp["DealsInMoreThanOneState"] = temp["DealsInMoreThanOneState"].replace("Yes", 1)

contracter_data = contracter_data.merge(temp[["ContractorId","DealsInMoreThanOneState"]], on="ContractorId", how="left")

temp["DealsInMoreThanOneState"].value_counts()
Out[101]:
0    1585
1     322
Name: DealsInMoreThanOneState, dtype: int64
In [102]:
temp = data[["ContractorId", "ApplicationDate","ApplicantId"]].groupby(["ContractorId","ApplicationDate"]).ApplicantId.nunique().reset_index()
temp.columns= ["ContractorId", "ApplicationDate","No_of_Applicants_on_same_Day"]
t1 = temp[["ContractorId", "No_of_Applicants_on_same_Day"]].groupby("ContractorId").min().reset_index()
t1.columns = ["ContractorId", "Minimum_No_of_Applicants_on_same_Day"]
contracter_data = contracter_data.merge(t1, on="ContractorId", how="left")

t1 = temp[["ContractorId", "No_of_Applicants_on_same_Day"]].groupby("ContractorId").max().reset_index()
t1.columns = ["ContractorId", "Maximum_No_of_Applicants_on_same_Day"]
contracter_data = contracter_data.merge(t1, on="ContractorId", how="left")

t1 = temp[["ContractorId", "No_of_Applicants_on_same_Day"]].groupby("ContractorId").mean().reset_index()
t1.columns = ["ContractorId", "Average_No_of_Applicants_on_same_Day"]
contracter_data = contracter_data.merge(t1, on="ContractorId", how="left")

Machine Learning Technique to identify Unique Contractors¶

In [103]:
from sklearn.ensemble import IsolationForest
contracter_data = contracter_data.fillna(0)
X = contracter_data.drop(columns=["ContractorId"])
clf = IsolationForest(n_estimators=500, random_state=0).fit(X)
contracter_data["Group"] = clf.predict(X)
In [104]:
contracter_data["Group"].value_counts()
Out[104]:
 1    1773
-1     134
Name: Group, dtype: int64

Contractors Different than Others¶

In [105]:
contracter_data[contracter_data["Group"] == -1].head(10)
Out[105]:
ContractorId TotalApplicants Approved Conditionally Approved Final - Canceled Final - Denied NetFundedAmount_Avg NetFundedAmount_Minimum NetFundedAmount_Maximum NetFundedAmount_Total Average_Applications_per_Day BusinessAge Applications_to_Approve_Ratio Applications_to_Denied_Ratio Over_all_approved_ration DealsInMoreThanOneState Minimum_No_of_Applicants_on_same_Day Maximum_No_of_Applicants_on_same_Day Average_No_of_Applicants_on_same_Day Group
1 35197 229 70 0 18 107 8744.263810 2990.32 23002.50 183629.54 6.939394 45 3.271429 2.140187 1.555556 0 2 17 6.939394 -1
3 39366 212 53 0 16 113 10187.096154 1987.42 22546.37 662161.25 2.436782 16 4.000000 1.876106 3.312500 1 1 8 2.436782 -1
4 39358 186 42 0 16 97 12501.569286 11024.64 23002.50 350043.94 2.818182 16 4.428571 1.917526 2.625000 1 1 10 2.818182 -1
5 38400 830 239 0 70 421 13129.280459 4994.67 30000.00 2573338.97 4.048780 89 3.472803 1.971496 2.685393 1 1 15 4.048780 -1
6 32274 1638 447 2 134 856 10846.945942 840.54 58974.24 10423915.05 6.275862 61 3.664430 1.913551 7.327869 1 1 29 6.275862 -1
7 37183 1512 385 0 124 806 11578.521060 1865.26 47505.00 9610172.48 6.379747 32 3.927273 1.875931 12.031250 1 1 21 6.379747 -1
9 38065 726 178 1 58 385 12365.044171 1813.71 47505.00 4921287.58 3.882353 26 4.078652 1.885714 6.846154 1 1 17 3.882353 -1
12 28694 493 122 0 33 267 10979.375288 2024.07 28887.49 1141855.03 3.100629 89 4.040984 1.846442 1.370787 1 1 11 3.100629 -1
14 39638 201 52 0 19 103 15124.319846 4600.50 32383.84 983080.79 2.115789 12 3.865385 1.951456 4.333333 1 1 8 2.115789 -1
15 15465 238 59 0 22 130 3970.601071 1728.87 19576.80 444707.32 2.034188 99 4.033898 1.830769 0.595960 0 1 5 2.034188 -1

Contractor (Majority Class)¶

In [106]:
contracter_data[contracter_data["Group"] == 1].head(10)
Out[106]:
ContractorId TotalApplicants Approved Conditionally Approved Final - Canceled Final - Denied NetFundedAmount_Avg NetFundedAmount_Minimum NetFundedAmount_Maximum NetFundedAmount_Total Average_Applications_per_Day BusinessAge Applications_to_Approve_Ratio Applications_to_Denied_Ratio Over_all_approved_ration DealsInMoreThanOneState Minimum_No_of_Applicants_on_same_Day Maximum_No_of_Applicants_on_same_Day Average_No_of_Applicants_on_same_Day Group
0 39390 103 28 0 9 53 11204.673333 8920.37 18503.21 67228.04 2.395349 15 3.678571 1.943396 1.866667 0 2 6 2.395349 1
2 36028 9 2 0 1 6 4236.306667 4108.92 4300.00 12708.92 1.800000 40 4.500000 1.500000 0.050000 0 1 2 1.800000 1
8 32244 79 15 0 10 43 15244.199444 10168.33 24627.50 274395.59 2.468750 61 5.266667 1.837209 0.245902 0 1 8 2.468750 1
10 30440 85 25 0 10 39 9091.752041 4432.12 23627.50 445495.85 1.847826 69 3.400000 2.179487 0.362319 0 1 6 1.847826 1
11 38363 151 44 0 10 80 14045.463600 5781.07 35330.99 702273.18 2.126761 89 3.431818 1.887500 0.494382 1 1 8 2.126761 1
13 39635 28 7 0 1 18 11046.048571 7809.42 24502.50 154644.68 1.400000 12 4.000000 1.555556 0.583333 0 1 2 1.400000 1
16 39547 113 39 0 14 53 9658.568837 1511.00 20243.42 415318.46 1.982456 16 2.897436 2.132075 2.437500 1 1 10 1.982456 1
21 37996 39 5 0 2 27 4073.372941 4862.78 6301.24 69247.34 2.294118 27 7.800000 1.444444 0.185185 0 1 8 2.294118 1
23 37421 38 7 0 4 21 8223.560000 13689.64 14128.52 41117.80 2.000000 31 5.428571 1.809524 0.225806 0 1 4 2.000000 1
24 38254 138 32 0 9 83 10578.869773 3435.00 23738.94 465470.27 1.943662 25 4.312500 1.662651 1.280000 1 1 10 1.943662 1
In [71]:
contracter_data.to_csv("Contractors_data.csv", index=False)
In [135]:
df = pd.DataFrame()
df["Feature"] = contracter_data[contracter_data["Group"] == -1].mean().index
df["Outlier_Mean"] = contracter_data[contracter_data["Group"] == -1].mean().values
df["Majority_Mean"] = contracter_data[contracter_data["Group"] == 1].mean().values
df["Outlier_Mean"] = round(df["Outlier_Mean"],2)
df
Out[135]:
Feature Outlier_Mean Majority_Mean
0 ContractorId 34659.01 35181.631698
1 TotalApplicants 330.49 17.745065
2 Approved 85.90 4.573040
3 Conditionally Approved 0.18 0.003384
4 Final - Canceled 28.52 1.545403
5 Final - Denied 171.28 9.280316
6 NetFundedAmount_Avg 14349.98 9389.537260
7 NetFundedAmount_Minimum 7840.27 4521.402482
8 NetFundedAmount_Maximum 31023.23 13201.230508
9 NetFundedAmount_Total 1204490.66 64766.695539
10 Average_Applications_per_Day 3.23 1.884910
11 BusinessAge 46.92 36.842076
12 Applications_to_Approve_Ratio 3.67 3.006531
13 Applications_to_Denied_Ratio 2.00 1.807478
14 Over_all_approved_ration 3.42 0.237806
15 DealsInMoreThanOneState 0.57 0.138748
16 Minimum_No_of_Applicants_on_same_Day 1.45 1.310773
17 Maximum_No_of_Applicants_on_same_Day 10.45 3.208122
18 Average_No_of_Applicants_on_same_Day 3.23 1.884910
19 Group -1.00 1.000000
In [136]:
df = pd.DataFrame()
df["Feature"] = contracter_data[contracter_data["Group"] == -1].max().index
df["Outlier_Maximum_Values"] = contracter_data[contracter_data["Group"] == -1].max().values
df["Majority_Maximum_Values"] = contracter_data[contracter_data["Group"] == 1].max().values
df["Outlier_Maximum_Values"] = round(df["Outlier_Maximum_Values"],2)
df
Out[136]:
Feature Outlier_Maximum_Values Majority_Maximum_Values
0 ContractorId 40482.00 40661.00
1 TotalApplicants 3249.00 189.00
2 Approved 803.00 53.00
3 Conditionally Approved 2.00 1.00
4 Final - Canceled 272.00 22.00
5 Final - Denied 1754.00 105.00
6 NetFundedAmount_Avg 58293.77 50000.00
7 NetFundedAmount_Minimum 58293.77 42966.38
8 NetFundedAmount_Maximum 72757.50 72757.50
9 NetFundedAmount_Total 16621294.60 1254797.56
10 Average_Applications_per_Day 11.90 5.00
11 BusinessAge 107.00 107.00
12 Applications_to_Approve_Ratio 8.00 24.00
13 Applications_to_Denied_Ratio 8.00 9.00
14 Over_all_approved_ration 44.00 7.50
15 DealsInMoreThanOneState 1.00 1.00
16 Minimum_No_of_Applicants_on_same_Day 8.00 5.00
17 Maximum_No_of_Applicants_on_same_Day 49.00 16.00
18 Average_No_of_Applicants_on_same_Day 11.90 5.00
19 Group -1.00 1.00
In [ ]:

Models With FICO and other Scores¶

In [149]:
columns = ["Mailing State", "EmploymentType", "AgeAtApplication", "StatedMonthlyIncome", "FinancialProduct",\
            "ApplicationDateDim", "ProjectType", "Property State", "RequestedAmount", "LoanTermInMonths", \
           "WhetherApplicantAvailLoanPreviously",
            "MonthlyPayment", "InterestRate", "AUSStatus", "LoanStatus", "AutoDecision", "SoftDecision", "HardDecision"]

sub_dataset = sub_dataset.fillna(0.0)

sub_dataset = sub_dataset[columns]

for i in ["Mailing State", "EmploymentType", "FinancialProduct", "ApplicationDateDim", "ProjectType", "Property State"]:
    LAYER = pd.get_dummies(sub_dataset[i], prefix=i+"_")
    sub_dataset[LAYER.columns] = LAYER
    sub_dataset = sub_dataset.drop(columns=[i])
In [152]:
dataset_ausstaus = sub_dataset[sub_dataset["AUSStatus"].isin(["Denied", "Approved"])][sub_dataset.columns]
Y_ausstatus = sub_dataset[sub_dataset["AUSStatus"].isin(["Denied", "Approved"])][["AUSStatus"]]
dataset_ausstaus = dataset_ausstaus.drop(columns=["AUSStatus", "LoanStatus", "AutoDecision", "SoftDecision", "HardDecision"])

dataset_loan = sub_dataset[sub_dataset["LoanStatus"].isin(["Final - Denied","Approved", "Final - Expired", "Final - Canceled"])][sub_dataset.columns]
Y_loan_status = sub_dataset[sub_dataset["LoanStatus"].isin(["Final - Denied","Approved", "Final - Expired", "Final - Canceled"])][["LoanStatus"]]
dataset_loan = dataset_loan.drop(columns=["AUSStatus", "LoanStatus", "AutoDecision", "SoftDecision", "HardDecision"])

dataset_auto_decision = sub_dataset[sub_dataset["AutoDecision"].isin(["Denied", "Approved"])][sub_dataset.columns]
Y_auto_decision = sub_dataset[sub_dataset["AutoDecision"].isin(["Denied", "Approved"])][["AutoDecision"]]
dataset_auto_decision = dataset_auto_decision.drop(columns=["AUSStatus", "LoanStatus", "AutoDecision", "SoftDecision", "HardDecision"])

dataset_soft_decision = sub_dataset[sub_dataset["SoftDecision"].isin(["Denied", "Approved"])][sub_dataset.columns]
Y_soft_decision = sub_dataset[sub_dataset["SoftDecision"].isin(["Denied", "Approved"])][["SoftDecision"]]
dataset_soft_decision = dataset_soft_decision.drop(columns=["AUSStatus", "LoanStatus", "AutoDecision", "SoftDecision", "HardDecision"])

dataset_hard_decision = sub_dataset[sub_dataset["HardDecision"].isin(["Denied", "Approved"])][sub_dataset.columns]
Y_hard_decision = sub_dataset[sub_dataset["HardDecision"].isin(["Denied", "Approved"])][["HardDecision"]]
dataset_hard_decision = dataset_hard_decision.drop(columns=["AUSStatus", "LoanStatus", "AutoDecision", "SoftDecision", "HardDecision"])

print("**************************************")
print("AUSStatus Prediction")
print("**************************************")
trainMLAlgorithm(dataset_ausstaus, Y_ausstatus)

print("**************************************")
print("Loan Status Prediction")
print("**************************************")
trainMLAlgorithm(dataset_loan, Y_loan_status)

print("**************************************")
print("Auto Decision Prediction")
print("**************************************")
trainMLAlgorithm(dataset_auto_decision, Y_auto_decision)

print("**************************************")
print("Soft Decision Prediction")
print("**************************************")
trainMLAlgorithm(dataset_soft_decision, Y_soft_decision)

print("**************************************")
print("Hard Decision Prediction")
print("**************************************")
trainMLAlgorithm(dataset_hard_decision, Y_hard_decision)

**************************************
AUSStatus Prediction
**************************************
Classification Report
              precision    recall  f1-score   support

    Approved       0.98      0.96      0.97      2697
      Denied       0.98      0.99      0.98      4369

    accuracy                           0.98      7066
   macro avg       0.98      0.97      0.98      7066
weighted avg       0.98      0.98      0.98      7066


Confusion Matrix
[[2600   97]
 [  64 4305]]

**************************************
Loan Status Prediction
**************************************
Classification Report
                  precision    recall  f1-score   support

        Approved       0.83      0.99      0.90      5886
Final - Canceled       0.28      0.14      0.19      1969
  Final - Denied       0.84      0.91      0.87     11821
 Final - Expired       0.34      0.19      0.25      2612

        accuracy                           0.78     22288
       macro avg       0.57      0.56      0.55     22288
    weighted avg       0.73      0.78      0.75     22288


Confusion Matrix
[[ 5818    27     0    41]
 [  474   275   932   288]
 [   27   418 10729   647]
 [  692   266  1151   503]]

**************************************
Auto Decision Prediction
**************************************
Classification Report
              precision    recall  f1-score   support

    Approved       0.97      0.95      0.96      1861
      Denied       0.99      0.99      0.99      7724

    accuracy                           0.98      9585
   macro avg       0.98      0.97      0.98      9585
weighted avg       0.98      0.98      0.98      9585


Confusion Matrix
[[1776   85]
 [  64 7660]]

**************************************
Soft Decision Prediction
**************************************
Classification Report
              precision    recall  f1-score   support

    Approved       0.94      0.87      0.90      2700
      Denied       0.97      0.99      0.98     11796

    accuracy                           0.97     14496
   macro avg       0.95      0.93      0.94     14496
weighted avg       0.97      0.97      0.97     14496


Confusion Matrix
[[ 2360   340]
 [  160 11636]]

**************************************
Hard Decision Prediction
**************************************
Classification Report
              precision    recall  f1-score   support

    Approved       0.99      1.00      1.00      6591
      Denied       1.00      0.18      0.31        50

    accuracy                           0.99      6641
   macro avg       1.00      0.59      0.65      6641
weighted avg       0.99      0.99      0.99      6641


Confusion Matrix
[[6591    0]
 [  41    9]]

Financial Product¶

In [242]:
benji = data[data["FinancialProduct"] == "Benji"][["ApplicantId", "ApplicationDate"]].groupby(["ApplicationDate"]).count().reset_index()
benji.head(5)
Out[242]:
ApplicationDate ApplicantId
0 2021-06-01 266
1 2021-06-02 273
2 2021-06-03 285
3 2021-06-04 262
4 2021-06-05 164
In [243]:
X, y = create_dataset(benji["ApplicantId"].values.reshape(-1, 1))
X_train, X_test = X[0:int(0.7*len(X))], X[int(0.7*len(X)):]
y_train, y_test = y[0:int(0.7*len(y))], y[int(0.7*len(y)):]

y_axis = applicant_per_month["ApplicationDate"].values
y_axis = y_axis[int(0.73*len(y_axis)):]

from sklearn.ensemble import RandomForestRegressor
model = RandomForestRegressor(n_estimators=500)
model.fit(X_train, y_train)
y_pred = model.predict(X_test)
score = model.score(X_test, y_test)

print('ROC AUC score: {:.2f}'.format(score))

rmse = math.sqrt(mean_squared_error(y_test, y_pred))
print("RMSE = ", rmse)
print("R2 Score = ", r2_score(y_test, y_pred))
print("MAE = ", mean_absolute_error(y_test, y_pred))

plt.figure(figsize=(20,5))
plt.plot(y_axis[0:50], y_test[0:50], '--gx', markersize = 5 , linewidth = 1, label="Actual - Number of Loan Applications for Benji")
plt.plot(y_axis[0:50], y_pred[0:50], '--rx', markersize = 5 , linewidth = 1, label="Predicted - Number of Loan Applications for Benji")
plt.legend()

fig = plt.figure(figsize=(16,8))
a = [i for i in range(len(y_test))]
L = 0
H = 500
plt.plot(y_axis[L:H], y_test[L:H], label='Actual - Number of Loan Applications for Benji')
plt.plot(y_axis[L:H], y_pred[L:H], label='Predicted - Number of Loan Applications for Benji')
plt.legend()
plt.ylabel("Number of Loan Applications for Benji")
plt.show()

ROC AUC score: 0.78
RMSE =  29.39147808309563
R2 Score =  0.7821113328248621
MAE =  23.044363636363638
In [244]:
print("Applications Next Month")
nextMonth(model, X)

Applications Next Month
Out[244]:
4715.0
In [245]:
LendingPoint = data[data["FinancialProduct"] == "LendingPoint"][["ApplicantId", "ApplicationDate"]].groupby(["ApplicationDate"]).count().reset_index()
LendingPoint.head(5)
Out[245]:
ApplicationDate ApplicantId
0 2021-06-01 145
1 2021-06-02 134
2 2021-06-03 139
3 2021-06-04 126
4 2021-06-05 99
In [246]:
X, y = create_dataset(LendingPoint["ApplicantId"].values.reshape(-1, 1))
X_train, X_test = X[0:int(0.7*len(X))], X[int(0.7*len(X)):]
y_train, y_test = y[0:int(0.7*len(y))], y[int(0.7*len(y)):]

y_axis = applicant_per_month["ApplicationDate"].values
y_axis = y_axis[int(0.73*len(y_axis)):]

from sklearn.ensemble import RandomForestRegressor
model = RandomForestRegressor(n_estimators=500)
model.fit(X_train, y_train)
y_pred = model.predict(X_test)
score = model.score(X_test, y_test)

print('ROC AUC score: {:.2f}'.format(score))

rmse = math.sqrt(mean_squared_error(y_test, y_pred))
print("RMSE = ", rmse)
print("R2 Score = ", r2_score(y_test, y_pred))
print("MAE = ", mean_absolute_error(y_test, y_pred))

plt.figure(figsize=(20,5))
plt.plot(y_axis[0:50], y_test[0:50], '--gx', markersize = 5 , linewidth = 1, label="Actual - Number of Loan Applications for LendingPoint")
plt.plot(y_axis[0:50], y_pred[0:50], '--rx', markersize = 5 , linewidth = 1, label="Predicted - Number of Loan Applications for LendingPoint")
plt.legend()

fig = plt.figure(figsize=(16,8))
a = [i for i in range(len(y_test))]
L = 0
H = 500
plt.plot(y_axis[L:H], y_test[L:H], label='Actual - Number of Loan Applications for LendingPoint')
plt.plot(y_axis[L:H], y_pred[L:H], label='Predicted - Number of Loan Applications for LendingPoint')
plt.legend()
plt.ylabel("Number of Loan Applications for LendingPoint")
plt.show()

ROC AUC score: 0.73
RMSE =  15.036411857403166
R2 Score =  0.7315878338308246
MAE =  11.599727272727272
In [247]:
print("Applications Next Month")
nextMonth(model, X)

Applications Next Month
Out[247]:
2193.0
In [ ]: